AIR Lab (AI Innovation & Research Lab) conducts fundamental and applied research in Multimodal Foundation Models, Autonomous Agentic Systems, Bio-Medical Intelligence, and Trustworthy AI.
Our research agenda spans foundational machine learning theories, multi-agent cognitive architectures, and transformative real-world applications.
Experience our latest foundation models and reasoning architectures directly inside your browser.
Observe test-time search, intermediate hypothesis backtracking, and verification kernels in real time.
Continuous-discrete tokenization enables zero-shot 3D spatial grounding, instance detection, and semantic segmentation in single-pass inference.
Benchmarked across standardized academic evaluation suites: MATH-500 / MMLU, GSM8K, and HumanEval.
| Model Architecture | MATH / MMLU (Reasoning) | GSM8K (Grade School Math) | HumanEval (Python Code) | Latency (TTFT) | Weights Status |
|---|
Peer-reviewed conference papers, journal articles, and open preprints advancing the science of artificial intelligence.
We believe in reproducible, open-weights AI. Explore our repositories, model checkpoints, and evaluation benchmarks.
A diverse collective of research scientists, doctoral candidates, engineers, and scholars.
Stay up to date with the latest laboratory breakthroughs, conference presentations, and upcoming guest talks.
We are actively looking for passionate researchers, doctoral students, and engineering interns to join our laboratory.
Focus on LLM Mechanistic Interpretability, Sparse Autoencoders, and Alignment.
Fully funded assistantships in Multimodal VLM, Autonomous Agents, and Bio-AI.
Hands-on mentorship on training distributed transformers on GPU superclusters.
Joint research programs, sponsored compute hours, and technology transfer.
Fill out the form below. Our admissions and collaboration committee reviews inquiries on a rolling basis.
